Presentation is loading. Please wait.

Presentation is loading. Please wait.

GPU Accelerated Vessel Segmentation Using Laplacian Eigenmaps Lin Cheng, Hyunsu Cho and Peter A. Yoon Trinity College.

Similar presentations


Presentation on theme: "GPU Accelerated Vessel Segmentation Using Laplacian Eigenmaps Lin Cheng, Hyunsu Cho and Peter A. Yoon Trinity College."— Presentation transcript:

1 GPU Accelerated Vessel Segmentation Using Laplacian Eigenmaps Lin Cheng, Hyunsu Cho and Peter A. Yoon Trinity College

2 Problem Image segmentation Partition pictures of vessels into segments

3 Local info embedded in high dimensional space Project local info onto low-dimensional plane Optimize the projection to preserve essential characteristics Cluster the projected data points into segments [1] Tziakos, Laskaris, and Fotopoulos

4 Segmentation process Build graph of local info Apply Laplace operator Solve optimization problem

5 Build graph of local info Store the resulting graph in a weight matrix Edges reflect variations among different regions (global variation)

6 Apply Laplace operator Form Laplacian matrix L = I – D 1/2 WD 1/2 encoding the Laplace operator. The operator formulates an optimization problem: Projections of well-connected nodes should also be tightly clustered.

7 Solve optimization problem Solutions to eigenvalue problem Ly = λy are optimal solutions If solutions are good, we can detect clusters

8 Characteristics of GPUs Massively parallel – lots of small cores (workers) Good for high-throughput, compute-bound tasks Separate memory space from main memory

9 Strategy: Reduce memory footprint On-GPU memory is limited Reduce memory usage and we can pack in more work into GPU

10 Strategy: Reduce memory footprint Weight matrix generation: Do not store intermediate results More entries can be calculated in parallel; 10x faster

11 Worker allocation

12 Strategy: use Lanczos method We need only a few smallest eigenvalues of L Lanczos method iteratively solve for the eigenvalues needed Takes 1/28 time of conventional method

13 Performance

14 Performance: vs. multicore CPUs CPU: two Intel® Xeon® E5-2620 GPU: one Nvidia Tesla® K20c

15 Acknowledgement Trinity College, Student Research Program Nvidia Corporation, CUDA Teaching Center Program


Download ppt "GPU Accelerated Vessel Segmentation Using Laplacian Eigenmaps Lin Cheng, Hyunsu Cho and Peter A. Yoon Trinity College."

Similar presentations


Ads by Google